2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
6 * Routing netlink socket interface: protocol independent part.
8 * Authors: Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
10 * This program is free software; you can redistribute it and/or
11 * modify it under the terms of the GNU General Public License
12 * as published by the Free Software Foundation; either version
13 * 2 of the License, or (at your option) any later version.
16 * Vitaly E. Lavrov RTA_OK arithmetics was wrong.
19 #include <linux/errno.h>
20 #include <linux/module.h>
21 #include <linux/types.h>
22 #include <linux/socket.h>
23 #include <linux/kernel.h>
24 #include <linux/timer.h>
25 #include <linux/string.h>
26 #include <linux/sockios.h>
27 #include <linux/net.h>
28 #include <linux/fcntl.h>
30 #include <linux/slab.h>
31 #include <linux/interrupt.h>
32 #include <linux/capability.h>
33 #include <linux/skbuff.h>
34 #include <linux/init.h>
35 #include <linux/security.h>
36 #include <linux/mutex.h>
37 #include <linux/if_addr.h>
38 #include <linux/nsproxy.h>
40 #include <asm/uaccess.h>
41 #include <asm/system.h>
42 #include <asm/string.h>
44 #include <linux/inet.h>
45 #include <linux/netdevice.h>
47 #include <net/protocol.h>
49 #include <net/route.h>
52 #include <net/pkt_sched.h>
53 #include <net/fib_rules.h>
54 #include <net/rtnetlink.h>
59 rtnl_dumpit_func dumpit;
62 static DEFINE_MUTEX(rtnl_mutex);
66 mutex_lock(&rtnl_mutex);
69 void __rtnl_unlock(void)
71 mutex_unlock(&rtnl_mutex);
74 void rtnl_unlock(void)
76 mutex_unlock(&rtnl_mutex);
80 int rtnl_trylock(void)
82 return mutex_trylock(&rtnl_mutex);
85 int rtattr_parse(struct rtattr *tb[], int maxattr, struct rtattr *rta, int len)
87 memset(tb, 0, sizeof(struct rtattr*)*maxattr);
89 while (RTA_OK(rta, len)) {
90 unsigned flavor = rta->rta_type;
91 if (flavor && flavor <= maxattr)
93 rta = RTA_NEXT(rta, len);
98 int __rtattr_parse_nested_compat(struct rtattr *tb[], int maxattr,
99 struct rtattr *rta, int len)
101 if (RTA_PAYLOAD(rta) < len)
103 if (RTA_PAYLOAD(rta) >= RTA_ALIGN(len) + sizeof(struct rtattr)) {
104 rta = RTA_DATA(rta) + RTA_ALIGN(len);
105 return rtattr_parse_nested(tb, maxattr, rta);
107 memset(tb, 0, sizeof(struct rtattr *) * maxattr);
111 static struct rtnl_link *rtnl_msg_handlers[NPROTO];
113 static inline int rtm_msgindex(int msgtype)
115 int msgindex = msgtype - RTM_BASE;
118 * msgindex < 0 implies someone tried to register a netlink
119 * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
120 * the message type has not been added to linux/rtnetlink.h
122 BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
127 static rtnl_doit_func rtnl_get_doit(int protocol, int msgindex)
129 struct rtnl_link *tab;
131 tab = rtnl_msg_handlers[protocol];
132 if (tab == NULL || tab[msgindex].doit == NULL)
133 tab = rtnl_msg_handlers[PF_UNSPEC];
135 return tab ? tab[msgindex].doit : NULL;
138 static rtnl_dumpit_func rtnl_get_dumpit(int protocol, int msgindex)
140 struct rtnl_link *tab;
142 tab = rtnl_msg_handlers[protocol];
143 if (tab == NULL || tab[msgindex].dumpit == NULL)
144 tab = rtnl_msg_handlers[PF_UNSPEC];
146 return tab ? tab[msgindex].dumpit : NULL;
150 * __rtnl_register - Register a rtnetlink message type
151 * @protocol: Protocol family or PF_UNSPEC
152 * @msgtype: rtnetlink message type
153 * @doit: Function pointer called for each request message
154 * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
156 * Registers the specified function pointers (at least one of them has
157 * to be non-NULL) to be called whenever a request message for the
158 * specified protocol family and message type is received.
160 * The special protocol family PF_UNSPEC may be used to define fallback
161 * function pointers for the case when no entry for the specific protocol
164 * Returns 0 on success or a negative error code.
166 int __rtnl_register(int protocol, int msgtype,
167 rtnl_doit_func doit, rtnl_dumpit_func dumpit)
169 struct rtnl_link *tab;
172 BUG_ON(protocol < 0 || protocol >= NPROTO);
173 msgindex = rtm_msgindex(msgtype);
175 tab = rtnl_msg_handlers[protocol];
177 tab = kcalloc(RTM_NR_MSGTYPES, sizeof(*tab), GFP_KERNEL);
181 rtnl_msg_handlers[protocol] = tab;
185 tab[msgindex].doit = doit;
188 tab[msgindex].dumpit = dumpit;
193 EXPORT_SYMBOL_GPL(__rtnl_register);
196 * rtnl_register - Register a rtnetlink message type
198 * Identical to __rtnl_register() but panics on failure. This is useful
199 * as failure of this function is very unlikely, it can only happen due
200 * to lack of memory when allocating the chain to store all message
201 * handlers for a protocol. Meant for use in init functions where lack
202 * of memory implies no sense in continueing.
204 void rtnl_register(int protocol, int msgtype,
205 rtnl_doit_func doit, rtnl_dumpit_func dumpit)
207 if (__rtnl_register(protocol, msgtype, doit, dumpit) < 0)
208 panic("Unable to register rtnetlink message handler, "
209 "protocol = %d, message type = %d\n",
213 EXPORT_SYMBOL_GPL(rtnl_register);
216 * rtnl_unregister - Unregister a rtnetlink message type
217 * @protocol: Protocol family or PF_UNSPEC
218 * @msgtype: rtnetlink message type
220 * Returns 0 on success or a negative error code.
222 int rtnl_unregister(int protocol, int msgtype)
226 BUG_ON(protocol < 0 || protocol >= NPROTO);
227 msgindex = rtm_msgindex(msgtype);
229 if (rtnl_msg_handlers[protocol] == NULL)
232 rtnl_msg_handlers[protocol][msgindex].doit = NULL;
233 rtnl_msg_handlers[protocol][msgindex].dumpit = NULL;
238 EXPORT_SYMBOL_GPL(rtnl_unregister);
241 * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
242 * @protocol : Protocol family or PF_UNSPEC
244 * Identical to calling rtnl_unregster() for all registered message types
245 * of a certain protocol family.
247 void rtnl_unregister_all(int protocol)
249 BUG_ON(protocol < 0 || protocol >= NPROTO);
251 kfree(rtnl_msg_handlers[protocol]);
252 rtnl_msg_handlers[protocol] = NULL;
255 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
257 static LIST_HEAD(link_ops);
260 * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
261 * @ops: struct rtnl_link_ops * to register
263 * The caller must hold the rtnl_mutex. This function should be used
264 * by drivers that create devices during module initialization. It
265 * must be called before registering the devices.
267 * Returns 0 on success or a negative error code.
269 int __rtnl_link_register(struct rtnl_link_ops *ops)
272 ops->dellink = unregister_netdevice;
274 list_add_tail(&ops->list, &link_ops);
278 EXPORT_SYMBOL_GPL(__rtnl_link_register);
281 * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
282 * @ops: struct rtnl_link_ops * to register
284 * Returns 0 on success or a negative error code.
286 int rtnl_link_register(struct rtnl_link_ops *ops)
291 err = __rtnl_link_register(ops);
296 EXPORT_SYMBOL_GPL(rtnl_link_register);
299 * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
300 * @ops: struct rtnl_link_ops * to unregister
302 * The caller must hold the rtnl_mutex.
304 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
306 struct net_device *dev, *n;
311 for_each_netdev_safe(net, dev, n) {
312 if (dev->rtnl_link_ops == ops) {
318 list_del(&ops->list);
321 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
324 * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
325 * @ops: struct rtnl_link_ops * to unregister
327 void rtnl_link_unregister(struct rtnl_link_ops *ops)
330 __rtnl_link_unregister(ops);
334 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
336 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
338 const struct rtnl_link_ops *ops;
340 list_for_each_entry(ops, &link_ops, list) {
341 if (!strcmp(ops->kind, kind))
347 static size_t rtnl_link_get_size(const struct net_device *dev)
349 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
355 size = nlmsg_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
356 nlmsg_total_size(strlen(ops->kind) + 1); /* IFLA_INFO_KIND */
359 /* IFLA_INFO_DATA + nested data */
360 size += nlmsg_total_size(sizeof(struct nlattr)) +
363 if (ops->get_xstats_size)
364 size += ops->get_xstats_size(dev); /* IFLA_INFO_XSTATS */
369 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
371 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
372 struct nlattr *linkinfo, *data;
375 linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
376 if (linkinfo == NULL)
379 if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
380 goto err_cancel_link;
381 if (ops->fill_xstats) {
382 err = ops->fill_xstats(skb, dev);
384 goto err_cancel_link;
386 if (ops->fill_info) {
387 data = nla_nest_start(skb, IFLA_INFO_DATA);
389 goto err_cancel_link;
390 err = ops->fill_info(skb, dev);
392 goto err_cancel_data;
393 nla_nest_end(skb, data);
396 nla_nest_end(skb, linkinfo);
400 nla_nest_cancel(skb, data);
402 nla_nest_cancel(skb, linkinfo);
407 static const int rtm_min[RTM_NR_FAMILIES] =
409 [RTM_FAM(RTM_NEWLINK)] = NLMSG_LENGTH(sizeof(struct ifinfomsg)),
410 [RTM_FAM(RTM_NEWADDR)] = NLMSG_LENGTH(sizeof(struct ifaddrmsg)),
411 [RTM_FAM(RTM_NEWROUTE)] = NLMSG_LENGTH(sizeof(struct rtmsg)),
412 [RTM_FAM(RTM_NEWRULE)] = NLMSG_LENGTH(sizeof(struct fib_rule_hdr)),
413 [RTM_FAM(RTM_NEWQDISC)] = NLMSG_LENGTH(sizeof(struct tcmsg)),
414 [RTM_FAM(RTM_NEWTCLASS)] = NLMSG_LENGTH(sizeof(struct tcmsg)),
415 [RTM_FAM(RTM_NEWTFILTER)] = NLMSG_LENGTH(sizeof(struct tcmsg)),
416 [RTM_FAM(RTM_NEWACTION)] = NLMSG_LENGTH(sizeof(struct tcamsg)),
417 [RTM_FAM(RTM_GETMULTICAST)] = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
418 [RTM_FAM(RTM_GETANYCAST)] = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
421 static const int rta_max[RTM_NR_FAMILIES] =
423 [RTM_FAM(RTM_NEWLINK)] = IFLA_MAX,
424 [RTM_FAM(RTM_NEWADDR)] = IFA_MAX,
425 [RTM_FAM(RTM_NEWROUTE)] = RTA_MAX,
426 [RTM_FAM(RTM_NEWRULE)] = FRA_MAX,
427 [RTM_FAM(RTM_NEWQDISC)] = TCA_MAX,
428 [RTM_FAM(RTM_NEWTCLASS)] = TCA_MAX,
429 [RTM_FAM(RTM_NEWTFILTER)] = TCA_MAX,
430 [RTM_FAM(RTM_NEWACTION)] = TCAA_MAX,
433 void __rta_fill(struct sk_buff *skb, int attrtype, int attrlen, const void *data)
436 int size = RTA_LENGTH(attrlen);
438 rta = (struct rtattr*)skb_put(skb, RTA_ALIGN(size));
439 rta->rta_type = attrtype;
441 memcpy(RTA_DATA(rta), data, attrlen);
442 memset(RTA_DATA(rta) + attrlen, 0, RTA_ALIGN(size) - size);
445 size_t rtattr_strlcpy(char *dest, const struct rtattr *rta, size_t size)
447 size_t ret = RTA_PAYLOAD(rta);
448 char *src = RTA_DATA(rta);
450 if (ret > 0 && src[ret - 1] == '\0')
453 size_t len = (ret >= size) ? size - 1 : ret;
454 memset(dest, 0, size);
455 memcpy(dest, src, len);
460 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned group, int echo)
462 struct sock *rtnl = net->rtnl;
465 NETLINK_CB(skb).dst_group = group;
467 atomic_inc(&skb->users);
468 netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
470 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
474 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
476 struct sock *rtnl = net->rtnl;
478 return nlmsg_unicast(rtnl, skb, pid);
481 int rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
482 struct nlmsghdr *nlh, gfp_t flags)
484 struct sock *rtnl = net->rtnl;
488 report = nlmsg_report(nlh);
490 return nlmsg_notify(rtnl, skb, pid, group, report, flags);
493 void rtnl_set_sk_err(struct net *net, u32 group, int error)
495 struct sock *rtnl = net->rtnl;
497 netlink_set_err(rtnl, 0, group, error);
500 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
505 mx = nla_nest_start(skb, RTA_METRICS);
509 for (i = 0; i < RTAX_MAX; i++) {
512 NLA_PUT_U32(skb, i+1, metrics[i]);
517 nla_nest_cancel(skb, mx);
521 return nla_nest_end(skb, mx);
524 return nla_nest_cancel(skb, mx);
527 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
528 u32 ts, u32 tsage, long expires, u32 error)
530 struct rta_cacheinfo ci = {
531 .rta_lastuse = jiffies_to_clock_t(jiffies - dst->lastuse),
532 .rta_used = dst->__use,
533 .rta_clntref = atomic_read(&(dst->__refcnt)),
541 ci.rta_expires = jiffies_to_clock_t(expires);
543 return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
546 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
548 static void set_operstate(struct net_device *dev, unsigned char transition)
550 unsigned char operstate = dev->operstate;
554 if ((operstate == IF_OPER_DORMANT ||
555 operstate == IF_OPER_UNKNOWN) &&
557 operstate = IF_OPER_UP;
560 case IF_OPER_DORMANT:
561 if (operstate == IF_OPER_UP ||
562 operstate == IF_OPER_UNKNOWN)
563 operstate = IF_OPER_DORMANT;
567 if (dev->operstate != operstate) {
568 write_lock_bh(&dev_base_lock);
569 dev->operstate = operstate;
570 write_unlock_bh(&dev_base_lock);
571 netdev_state_change(dev);
575 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
576 struct net_device_stats *b)
578 a->rx_packets = b->rx_packets;
579 a->tx_packets = b->tx_packets;
580 a->rx_bytes = b->rx_bytes;
581 a->tx_bytes = b->tx_bytes;
582 a->rx_errors = b->rx_errors;
583 a->tx_errors = b->tx_errors;
584 a->rx_dropped = b->rx_dropped;
585 a->tx_dropped = b->tx_dropped;
587 a->multicast = b->multicast;
588 a->collisions = b->collisions;
590 a->rx_length_errors = b->rx_length_errors;
591 a->rx_over_errors = b->rx_over_errors;
592 a->rx_crc_errors = b->rx_crc_errors;
593 a->rx_frame_errors = b->rx_frame_errors;
594 a->rx_fifo_errors = b->rx_fifo_errors;
595 a->rx_missed_errors = b->rx_missed_errors;
597 a->tx_aborted_errors = b->tx_aborted_errors;
598 a->tx_carrier_errors = b->tx_carrier_errors;
599 a->tx_fifo_errors = b->tx_fifo_errors;
600 a->tx_heartbeat_errors = b->tx_heartbeat_errors;
601 a->tx_window_errors = b->tx_window_errors;
603 a->rx_compressed = b->rx_compressed;
604 a->tx_compressed = b->tx_compressed;
607 static inline size_t if_nlmsg_size(const struct net_device *dev)
609 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
610 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
611 + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
612 + nla_total_size(sizeof(struct rtnl_link_ifmap))
613 + nla_total_size(sizeof(struct rtnl_link_stats))
614 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
615 + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
616 + nla_total_size(4) /* IFLA_TXQLEN */
617 + nla_total_size(4) /* IFLA_WEIGHT */
618 + nla_total_size(4) /* IFLA_MTU */
619 + nla_total_size(4) /* IFLA_LINK */
620 + nla_total_size(4) /* IFLA_MASTER */
621 + nla_total_size(1) /* IFLA_OPERSTATE */
622 + nla_total_size(1) /* IFLA_LINKMODE */
623 + rtnl_link_get_size(dev); /* IFLA_LINKINFO */
626 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
627 int type, u32 pid, u32 seq, u32 change,
630 struct ifinfomsg *ifm;
631 struct nlmsghdr *nlh;
633 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
637 ifm = nlmsg_data(nlh);
638 ifm->ifi_family = AF_UNSPEC;
640 ifm->ifi_type = dev->type;
641 ifm->ifi_index = dev->ifindex;
642 ifm->ifi_flags = dev_get_flags(dev);
643 ifm->ifi_change = change;
645 NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
646 NLA_PUT_U32(skb, IFLA_TXQLEN, dev->tx_queue_len);
647 NLA_PUT_U8(skb, IFLA_OPERSTATE,
648 netif_running(dev) ? dev->operstate : IF_OPER_DOWN);
649 NLA_PUT_U8(skb, IFLA_LINKMODE, dev->link_mode);
650 NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
652 if (dev->ifindex != dev->iflink)
653 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
656 NLA_PUT_U32(skb, IFLA_MASTER, dev->master->ifindex);
658 if (dev->qdisc_sleeping)
659 NLA_PUT_STRING(skb, IFLA_QDISC, dev->qdisc_sleeping->ops->id);
662 struct rtnl_link_ifmap map = {
663 .mem_start = dev->mem_start,
664 .mem_end = dev->mem_end,
665 .base_addr = dev->base_addr,
668 .port = dev->if_port,
670 NLA_PUT(skb, IFLA_MAP, sizeof(map), &map);
674 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
675 NLA_PUT(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast);
678 if (dev->get_stats) {
679 struct net_device_stats *stats = dev->get_stats(dev);
683 attr = nla_reserve(skb, IFLA_STATS,
684 sizeof(struct rtnl_link_stats));
686 goto nla_put_failure;
688 copy_rtnl_link_stats(nla_data(attr), stats);
692 if (dev->rtnl_link_ops) {
693 if (rtnl_link_fill(skb, dev) < 0)
694 goto nla_put_failure;
697 return nlmsg_end(skb, nlh);
700 nlmsg_cancel(skb, nlh);
704 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
706 struct net *net = skb->sk->sk_net;
708 int s_idx = cb->args[0];
709 struct net_device *dev;
711 if (net != &init_net)
715 for_each_netdev(net, dev) {
718 if (rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
719 NETLINK_CB(cb->skb).pid,
720 cb->nlh->nlmsg_seq, 0, NLM_F_MULTI) <= 0)
730 const struct nla_policy ifla_policy[IFLA_MAX+1] = {
731 [IFLA_IFNAME] = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
732 [IFLA_ADDRESS] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
733 [IFLA_BROADCAST] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
734 [IFLA_MAP] = { .len = sizeof(struct rtnl_link_ifmap) },
735 [IFLA_MTU] = { .type = NLA_U32 },
736 [IFLA_TXQLEN] = { .type = NLA_U32 },
737 [IFLA_WEIGHT] = { .type = NLA_U32 },
738 [IFLA_OPERSTATE] = { .type = NLA_U8 },
739 [IFLA_LINKMODE] = { .type = NLA_U8 },
740 [IFLA_NET_NS_PID] = { .type = NLA_U32 },
743 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
744 [IFLA_INFO_KIND] = { .type = NLA_STRING },
745 [IFLA_INFO_DATA] = { .type = NLA_NESTED },
748 static struct net *get_net_ns_by_pid(pid_t pid)
750 struct task_struct *tsk;
753 /* Lookup the network namespace */
754 net = ERR_PTR(-ESRCH);
756 tsk = find_task_by_vpid(pid);
758 struct nsproxy *nsproxy;
759 nsproxy = task_nsproxy(tsk);
761 net = get_net(nsproxy->net_ns);
767 static int do_setlink(struct net_device *dev, struct ifinfomsg *ifm,
768 struct nlattr **tb, char *ifname, int modified)
770 int send_addr_notify = 0;
773 if (tb[IFLA_NET_NS_PID]) {
775 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
780 err = dev_change_net_namespace(dev, net, ifname);
788 struct rtnl_link_ifmap *u_map;
791 if (!dev->set_config) {
796 if (!netif_device_present(dev)) {
801 u_map = nla_data(tb[IFLA_MAP]);
802 k_map.mem_start = (unsigned long) u_map->mem_start;
803 k_map.mem_end = (unsigned long) u_map->mem_end;
804 k_map.base_addr = (unsigned short) u_map->base_addr;
805 k_map.irq = (unsigned char) u_map->irq;
806 k_map.dma = (unsigned char) u_map->dma;
807 k_map.port = (unsigned char) u_map->port;
809 err = dev->set_config(dev, &k_map);
816 if (tb[IFLA_ADDRESS]) {
820 if (!dev->set_mac_address) {
825 if (!netif_device_present(dev)) {
830 len = sizeof(sa_family_t) + dev->addr_len;
831 sa = kmalloc(len, GFP_KERNEL);
836 sa->sa_family = dev->type;
837 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
839 err = dev->set_mac_address(dev, sa);
843 send_addr_notify = 1;
848 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
855 * Interface selected by interface index but interface
856 * name provided implies that a name change has been
859 if (ifm->ifi_index > 0 && ifname[0]) {
860 err = dev_change_name(dev, ifname);
866 if (tb[IFLA_BROADCAST]) {
867 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
868 send_addr_notify = 1;
871 if (ifm->ifi_flags || ifm->ifi_change) {
872 unsigned int flags = ifm->ifi_flags;
874 /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
876 flags = (flags & ifm->ifi_change) |
877 (dev->flags & ~ifm->ifi_change);
878 dev_change_flags(dev, flags);
882 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
884 if (tb[IFLA_OPERSTATE])
885 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
887 if (tb[IFLA_LINKMODE]) {
888 write_lock_bh(&dev_base_lock);
889 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
890 write_unlock_bh(&dev_base_lock);
896 if (err < 0 && modified && net_ratelimit())
897 printk(KERN_WARNING "A link change request failed with "
898 "some changes comitted already. Interface %s may "
899 "have been left with an inconsistent configuration, "
900 "please check.\n", dev->name);
902 if (send_addr_notify)
903 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
907 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
909 struct net *net = skb->sk->sk_net;
910 struct ifinfomsg *ifm;
911 struct net_device *dev;
913 struct nlattr *tb[IFLA_MAX+1];
914 char ifname[IFNAMSIZ];
916 if (net != &init_net)
919 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
924 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
929 ifm = nlmsg_data(nlh);
930 if (ifm->ifi_index > 0)
931 dev = dev_get_by_index(net, ifm->ifi_index);
932 else if (tb[IFLA_IFNAME])
933 dev = dev_get_by_name(net, ifname);
942 if (tb[IFLA_ADDRESS] &&
943 nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
946 if (tb[IFLA_BROADCAST] &&
947 nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
950 err = do_setlink(dev, ifm, tb, ifname, 0);
957 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
959 struct net *net = skb->sk->sk_net;
960 const struct rtnl_link_ops *ops;
961 struct net_device *dev;
962 struct ifinfomsg *ifm;
963 char ifname[IFNAMSIZ];
964 struct nlattr *tb[IFLA_MAX+1];
967 if (net != &init_net)
970 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
975 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
977 ifm = nlmsg_data(nlh);
978 if (ifm->ifi_index > 0)
979 dev = __dev_get_by_index(net, ifm->ifi_index);
980 else if (tb[IFLA_IFNAME])
981 dev = __dev_get_by_name(net, ifname);
988 ops = dev->rtnl_link_ops;
996 struct net_device *rtnl_create_link(struct net *net, char *ifname,
997 const struct rtnl_link_ops *ops, struct nlattr *tb[])
1000 struct net_device *dev;
1003 dev = alloc_netdev(ops->priv_size, ifname, ops->setup);
1007 if (strchr(dev->name, '%')) {
1008 err = dev_alloc_name(dev, dev->name);
1014 dev->rtnl_link_ops = ops;
1017 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
1018 if (tb[IFLA_ADDRESS])
1019 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
1020 nla_len(tb[IFLA_ADDRESS]));
1021 if (tb[IFLA_BROADCAST])
1022 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
1023 nla_len(tb[IFLA_BROADCAST]));
1024 if (tb[IFLA_TXQLEN])
1025 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1026 if (tb[IFLA_OPERSTATE])
1027 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1028 if (tb[IFLA_LINKMODE])
1029 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1036 return ERR_PTR(err);
1039 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1041 struct net *net = skb->sk->sk_net;
1042 const struct rtnl_link_ops *ops;
1043 struct net_device *dev;
1044 struct ifinfomsg *ifm;
1045 char kind[MODULE_NAME_LEN];
1046 char ifname[IFNAMSIZ];
1047 struct nlattr *tb[IFLA_MAX+1];
1048 struct nlattr *linkinfo[IFLA_INFO_MAX+1];
1051 if (net != &init_net)
1057 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1061 if (tb[IFLA_IFNAME])
1062 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1066 ifm = nlmsg_data(nlh);
1067 if (ifm->ifi_index > 0)
1068 dev = __dev_get_by_index(net, ifm->ifi_index);
1070 dev = __dev_get_by_name(net, ifname);
1074 if (tb[IFLA_LINKINFO]) {
1075 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
1076 tb[IFLA_LINKINFO], ifla_info_policy);
1080 memset(linkinfo, 0, sizeof(linkinfo));
1082 if (linkinfo[IFLA_INFO_KIND]) {
1083 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
1084 ops = rtnl_link_ops_get(kind);
1091 struct nlattr *attr[ops ? ops->maxtype + 1 : 0], **data = NULL;
1094 if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
1095 err = nla_parse_nested(attr, ops->maxtype,
1096 linkinfo[IFLA_INFO_DATA],
1102 if (ops->validate) {
1103 err = ops->validate(tb, data);
1112 if (nlh->nlmsg_flags & NLM_F_EXCL)
1114 if (nlh->nlmsg_flags & NLM_F_REPLACE)
1117 if (linkinfo[IFLA_INFO_DATA]) {
1118 if (!ops || ops != dev->rtnl_link_ops ||
1122 err = ops->changelink(dev, tb, data);
1128 return do_setlink(dev, ifm, tb, ifname, modified);
1131 if (!(nlh->nlmsg_flags & NLM_F_CREATE))
1134 if (ifm->ifi_index || ifm->ifi_flags || ifm->ifi_change)
1136 if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
1143 request_module("rtnl-link-%s", kind);
1145 ops = rtnl_link_ops_get(kind);
1154 snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
1156 dev = rtnl_create_link(net, ifname, ops, tb);
1160 else if (ops->newlink)
1161 err = ops->newlink(dev, tb, data);
1163 err = register_netdevice(dev);
1165 if (err < 0 && !IS_ERR(dev))
1171 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
1173 struct net *net = skb->sk->sk_net;
1174 struct ifinfomsg *ifm;
1175 struct nlattr *tb[IFLA_MAX+1];
1176 struct net_device *dev = NULL;
1177 struct sk_buff *nskb;
1180 if (net != &init_net)
1183 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1187 ifm = nlmsg_data(nlh);
1188 if (ifm->ifi_index > 0) {
1189 dev = dev_get_by_index(net, ifm->ifi_index);
1195 nskb = nlmsg_new(if_nlmsg_size(dev), GFP_KERNEL);
1201 err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).pid,
1202 nlh->nlmsg_seq, 0, 0);
1204 /* -EMSGSIZE implies BUG in if_nlmsg_size */
1205 WARN_ON(err == -EMSGSIZE);
1209 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).pid);
1216 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
1218 struct net *net = skb->sk->sk_net;
1220 int s_idx = cb->family;
1222 if (net != &init_net)
1227 for (idx=1; idx<NPROTO; idx++) {
1228 int type = cb->nlh->nlmsg_type-RTM_BASE;
1229 if (idx < s_idx || idx == PF_PACKET)
1231 if (rtnl_msg_handlers[idx] == NULL ||
1232 rtnl_msg_handlers[idx][type].dumpit == NULL)
1235 memset(&cb->args[0], 0, sizeof(cb->args));
1236 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
1244 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned change)
1246 struct sk_buff *skb;
1249 skb = nlmsg_new(if_nlmsg_size(dev), GFP_KERNEL);
1253 err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0);
1255 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
1256 WARN_ON(err == -EMSGSIZE);
1260 err = rtnl_notify(skb, &init_net, 0, RTNLGRP_LINK, NULL, GFP_KERNEL);
1263 rtnl_set_sk_err(&init_net, RTNLGRP_LINK, err);
1266 /* Protected by RTNL sempahore. */
1267 static struct rtattr **rta_buf;
1268 static int rtattr_max;
1270 /* Process one rtnetlink message. */
1272 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
1274 struct net *net = skb->sk->sk_net;
1275 rtnl_doit_func doit;
1282 type = nlh->nlmsg_type;
1288 /* All the messages must have at least 1 byte length */
1289 if (nlh->nlmsg_len < NLMSG_LENGTH(sizeof(struct rtgenmsg)))
1292 family = ((struct rtgenmsg*)NLMSG_DATA(nlh))->rtgen_family;
1293 if (family >= NPROTO)
1294 return -EAFNOSUPPORT;
1299 if (kind != 2 && security_netlink_recv(skb, CAP_NET_ADMIN))
1302 if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
1304 rtnl_dumpit_func dumpit;
1306 dumpit = rtnl_get_dumpit(family, type);
1312 err = netlink_dump_start(rtnl, skb, nlh, dumpit, NULL);
1317 memset(rta_buf, 0, (rtattr_max * sizeof(struct rtattr *)));
1319 min_len = rtm_min[sz_idx];
1320 if (nlh->nlmsg_len < min_len)
1323 if (nlh->nlmsg_len > min_len) {
1324 int attrlen = nlh->nlmsg_len - NLMSG_ALIGN(min_len);
1325 struct rtattr *attr = (void*)nlh + NLMSG_ALIGN(min_len);
1327 while (RTA_OK(attr, attrlen)) {
1328 unsigned flavor = attr->rta_type;
1330 if (flavor > rta_max[sz_idx])
1332 rta_buf[flavor-1] = attr;
1334 attr = RTA_NEXT(attr, attrlen);
1338 doit = rtnl_get_doit(family, type);
1342 return doit(skb, nlh, (void *)&rta_buf[0]);
1345 static void rtnetlink_rcv(struct sk_buff *skb)
1348 netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
1352 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
1354 struct net_device *dev = ptr;
1356 if (dev->nd_net != &init_net)
1360 case NETDEV_UNREGISTER:
1361 rtmsg_ifinfo(RTM_DELLINK, dev, ~0U);
1363 case NETDEV_REGISTER:
1364 rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U);
1368 rtmsg_ifinfo(RTM_NEWLINK, dev, IFF_UP|IFF_RUNNING);
1371 case NETDEV_GOING_DOWN:
1374 rtmsg_ifinfo(RTM_NEWLINK, dev, 0);
1380 static struct notifier_block rtnetlink_dev_notifier = {
1381 .notifier_call = rtnetlink_event,
1385 static int rtnetlink_net_init(struct net *net)
1388 sk = netlink_kernel_create(net, NETLINK_ROUTE, RTNLGRP_MAX,
1389 rtnetlink_rcv, &rtnl_mutex, THIS_MODULE);
1393 /* Don't hold an extra reference on the namespace */
1394 put_net(sk->sk_net);
1399 static void rtnetlink_net_exit(struct net *net)
1401 struct sock *sk = net->rtnl;
1403 /* At the last minute lie and say this is a socket for the
1404 * initial network namespace. So the socket will be safe to
1407 sk->sk_net = get_net(&init_net);
1413 static struct pernet_operations rtnetlink_net_ops = {
1414 .init = rtnetlink_net_init,
1415 .exit = rtnetlink_net_exit,
1418 void __init rtnetlink_init(void)
1423 for (i = 0; i < ARRAY_SIZE(rta_max); i++)
1424 if (rta_max[i] > rtattr_max)
1425 rtattr_max = rta_max[i];
1426 rta_buf = kmalloc(rtattr_max * sizeof(struct rtattr *), GFP_KERNEL);
1428 panic("rtnetlink_init: cannot allocate rta_buf\n");
1430 if (register_pernet_subsys(&rtnetlink_net_ops))
1431 panic("rtnetlink_init: cannot initialize rtnetlink\n");
1433 netlink_set_nonroot(NETLINK_ROUTE, NL_NONROOT_RECV);
1434 register_netdevice_notifier(&rtnetlink_dev_notifier);
1436 rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink, rtnl_dump_ifinfo);
1437 rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL);
1438 rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL);
1439 rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL);
1441 rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all);
1442 rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all);
1445 EXPORT_SYMBOL(__rta_fill);
1446 EXPORT_SYMBOL(rtattr_strlcpy);
1447 EXPORT_SYMBOL(rtattr_parse);
1448 EXPORT_SYMBOL(__rtattr_parse_nested_compat);
1449 EXPORT_SYMBOL(rtnetlink_put_metrics);
1450 EXPORT_SYMBOL(rtnl_lock);
1451 EXPORT_SYMBOL(rtnl_trylock);
1452 EXPORT_SYMBOL(rtnl_unlock);
1453 EXPORT_SYMBOL(rtnl_unicast);
1454 EXPORT_SYMBOL(rtnl_notify);
1455 EXPORT_SYMBOL(rtnl_set_sk_err);
1456 EXPORT_SYMBOL(rtnl_create_link);
1457 EXPORT_SYMBOL(ifla_policy);